TypingHand / 鍵筆

Real Figures · Real Talk · Real Test

A Hong Kong 80s working man, amplifying output with AI in spare hours, bearing neutral witness to the AI tide.

📊

Real Figures

Data-driven intelligence. Weekly dashboard covering AI cultural penetration, crypto convergence, and stock evaluation.

✍️

Real Talk

The human behind the data. Career reflections, technology commentary, life observations.

🔬

Real Test

One-off experiments. Reproducing papers, building tools, testing ideas.

📖

Real Read

Technique explainers. Deep dives into AI papers, concepts, and methods — from first principles.

🤖 AI Era Observer

View all →
Free Preview
Issue #7 · June 28, 2026

AI Era Observer — 2026-06-28

Narration-of-Thought (NoT) inference-time scaffolding, intent-aware training for LLM safety, speculative decoding accelerates LLM inference. Weekly AI ecosystem intelligence.

Free Preview
Issue #6 · June 21, 2026

AI Era Observer — 2026-06-21

Medical multi-agent system DeepRoot, multilingual agent trajectory dataset, LLM knowledge conflict resolution. Weekly AI ecosystem intelligence.

Free Preview
Issue #5 · June 14, 2026

AI Era Observer — 2026-06-14

Multi-source AI ecosystem tracking — from arXiv papers to Hacker News discussions, GitHub repos to HuggingFace models. Weekly publication, data-driven neutral AI trend observation.

📊 Weekly Dashboard

View all →
Week 26 · June 28, 2026

Weekly Real-Figures Dashboard — 28 June, 2026 (week 26)

Three independent data snapshots: AI in news discourse, crypto convergence indicators, and factor-based stock picks.

weekly-dashboardreal-figuresdata
Week 25 · June 21, 2026

Weekly Real-Figures Dashboard — 21 June, 2026 (week 25)

Three independent data snapshots: AI in news discourse, crypto convergence indicators, and factor-based stock picks.

weekly-dashboardreal-figuresdata

🔬 Real Test — Latest Experiments

View all →
May 25, 2026

AI Model Face-Off 2026-05: 18 Models Put Through 7 Rounds

A benchmark pitting 18 mainstream LLMs across 3 weight classes through 7 rounds — from factual accuracy to classical Chinese prose rewriting.

AI modelsbenchmarkLLMface-off

✍️ Real Talk — Latest Essays

View all →
June 6, 2026

Three Limitations of AI Chatbots — When Search Engines Became Answer Machines

AI chatbots are just the entry point. From token-by-token generation to stale training data and outsourced thinking — three blind spots you may not have noticed, and why AI agents are the logical next step.

AI learningAI ChatbotLLMRAGAI AgentHong Kongreal talk
May 30, 2026

Learning AI After Work — A Hands-On Review of Seven Free Tools

A Hong Kong office worker's real journey of self-learning AI after hours — honest comparisons of seven free AI tools, from POE to Gemini.

AI learningfree toolscareer developmentHong Kong workersAI tool comparisonself-improvementreal-talk
May 19, 2026

Standing Before the AI Wave — One HK Working Man's Choice

From the Industrial Revolution to the AI revolution, a personal reflection from a Hong Kong worker born in the 80s. Why learn AI, why build a website — neutral witness, no signals to sell.

AI learningHong Kongcareer reflection80s-borntechnology wavesreal talk